Hi , i am building Transparent and Controllable AI reasoning systems.
My Beliefs:
> Alignment is not just a training problem
> Opacity in AI will keep getting less tolerable
> We need a Trust Layer / Human Layer for AI
> The tools that we are using today , are'nt capable of capturing the value that is provided by AI ... YET
(will keep evolving these beliefs)
Currently solving this problem by working on https://t.co/dxLqitWmQ1
Hot take: you don't need a co-founder anymore.
AI lets solo founders 3x themselves covering product, tech, and sales well enough to get to first customers and prove the thing actually works.
Hussein Yahfoufi, solo founder and CEO of Eventship and Techstars All-Star Mentor, breaks down why the startup gospel is changing, and what actually replaces the co-founder when things get hard.
https://t.co/4p6RujMy1v
Everyone on 100$ and 200$ plans of codex.
Isn't it better now to just refer a friend(buy a cheap plan yourself)
For a rate limits reset 4 times a month.
Than buying another 100/200 $ subscription.
We heard you wanted to use Codex rate limit resets on your own time.
Starting today, we’re rolling out the ability to save rate limit resets to use later.
We’re starting Go, Plus, Pro, and Business users with one free reset:
I understand why it must sound like that.
But we are taking a entirely different approach to trust and transparency.
Governance platforms usually miss where the trust is lost. Especially for complex work.
We believe it should be ingrained in the system of the knowledge workers who work with AI from the ground up. Further, we believe that complex problems have a very branched underlying structure, which does not fit present AI tools(as they force linearity)
We are building a canvas where all of the knowledge work is done on a graph, and you can delegate agents to parts of the graph and work alongside them. Let them run experiments, branch through approaches, synthesize where needed.
The actual edge is the underlying context architecture of it, which understands your reasoning patterns(which is essential for knowledge work).
The system shows how it arrived to a solution, over evidence graphs, what assumptions it took and what were the conclusions.
This take different shapes according to the specific knowledge workers or organisation
, as a researcher might have a different set of primitives, than a consultant. We cater it to the end user.
It’s even more than that.
Even if I send the exact same prompt at the exact same time and branch into 2 different timelines where everything was same until I sent the prompt.
External effects like other users can change the hidden path: routing, queueing, batching, GPU scheduling, sampling state, etc.
From my side, nothing changed.
From the system’s side, the computation wasn’t identical.
https://t.co/iwIhw7GjFT by the way this is what I'm referencing. People took offense or told me skill issues. Models just behave differently depending on time of day, months of the year, and potentially what year it is.